The effects of preventing a COVID-19 health crisis have had unintended consequences on domestic abuse (DA) victimization. We contribute to the literature on domestic abuse in lockdown by providing insight on how changing patterns of domestic abuse can explain differences in magnitudes reported across studies. We examine the patterns of domestic abuse during the COVID-19 lockdown in Greater London and find that the lockdown changed the nature of reporting and the type of relationship the abuse occurs within. While abuse by current partners as well as family members increased on average by 8.1% and 17.1% respectively over the lockdown period, abuse by ex-partners declined by 11.4%. These findings show that reporting the average change in domestic abuse during lockdown can be misleading when designing a policy response. Moreover, we show that all the increase in domestic abuse calls is driven by third party reporting, particularly evident in areas with high density. This suggests that under reporting is present in the lockdown, particularly in households where the abuse cannot be reported by an outsider. Although these findings pertain to the COVID-19 lockdown, they also highlight the role that victim exposure and proximity has in affecting domestic abuse.

We compare predictions from a conventional protocol‐based approach to risk assessment with those based on a machine‐learning approach. We first show that the conventional predictions are less accurate than, and have similar rates of negative prediction error as, a simple Bayes classifier that makes use of only the base failure rate. Machine‐learning algorithms based on the underlying risk assessment questionnaire do better under the assumption that negative prediction errors are more costly than positive prediction errors. Machine‐learning models based on two‐year criminal histories do even better. Indeed, adding the protocol‐based features to the criminal histories adds little to the predictive adequacy of the model. We suggest using the predictions based on criminal histories to prioritize incoming calls for service, and devising a more sensitive instrument to distinguish true from false positives that result from this initial screening.

No abstract